On Deep Steward
I do not mean to alarm, but there may be only a single Caroirendephindisplar remaining in a certain pond in the centre of Rotterdam. Close to extinction, too, is the Gaeouthescriably, of which only one specimen seemed to remain on the day I visited in August of this year. The Unisess was practically thriving: seven were reported that morning, and for that I am thankful.
I don’t know what these species are – not necessarily plant or animal, they are more basic; more first-principle than that. Their taxonomies are more basic even than morphology, certainly not the fiendish implications of culture: these objects are defined by movements, by speed, in the uncanny manner we might once have defined threats detected in the corner of our human eye – not ‘mountain lion’, but ‘fast & shadowy & behind me’ or ‘sniff, sweat – lunge: run’.
That lonely Caroirendephindisplar has been detected not by the corner of a human eye but by Deep Steward, a machine that watched the pond, detecting, inventing, learning, dividing the world into a taxonomy based on simple rules. Unlike most artificial intelligences, Deep Steward has not been fed a baseline of information: no series of water flora, or 100s of turtles, to give it something to go on. So it’s making everything up as it goes along, like an alien eye landing in Rotterdam for the first time. This first outing for the Steward, as part of the Neuhaus exhibition at Het Nieuwe Instituut, is something of an experiment for DS and its builders, Amsterdam-based artist Theun Karelse and Los Angeles colleague Ian Ingram. It’s an idea; a launchpad.
Karelse and Ingram began to think about Deep Steward after a fateful trip to a Finnish forest. As Ingram writes, ’We pointed a camera into the landscape of arctic Finland – full of lichen-covered rocks and twisted birch trees – and asked an AI to tell us what it saw there. It told us it saw snowmobiles.’
The danger of AI is never external: it is the failings of humankind in our understanding of, and care for, both each other and the non-human. Can we, perhaps, see those snowmobiles as the clearest statement of a profound human failure: the projection of landscape over environment? If landscape is the tool through which humans mediate their relationships to place and time, then landscape art is so often made versus the landscape – not about kin, but conquering; a celebration of the distance between we two. We (at least in the Anglo-American west) frequently see the landscape as territory, not terroir. As many have pointed out in recent years, it is this romantic idea of the human / nonhuman dichotomy that is at the root of our current crises.
Ian Ingram writes of attempting to create, in opposition to this, “ a parliament of AIs”. (The inspiration for this phrase is from Bruno Latour’s “parliament of things”, but perhaps could be likened, also, to Alan Moore’s ‘Parliament of Trees’ – a concept that seems closer to truth than even magus Moore imagined.) How do we give an AI the opportunity to know, not just about ‘trees’ or ‘reeds’, but about its local trees and specific clusters of lichen? How do we at least attempt to partially eliminate human prejudices and dichotomies from an AI’s early education?
I’m fascinated by Deep Steward – not so much as a piece of tech, of which it is an aesthetically pleasing and even quite funny example: its nomenclature, its excited grabs. (Each slightly different movement or appearance is circled, a line drawn to the side of the screen, where it is given or attached to a name, the way a trainspotter might ‘Ooh!’ at a new number or, perhaps more appropriately, the scribbled ‘Wow!’ of the famed Wow! Signal). No, it’s fascinating as a critique of our own taxonomies and the inherent prejudices of culture, of history, of ‘the way things are done’.
There is a turtle in the pond in Rotterdam. Deep Steward doesn’t really care about the turtle, at least not most of the time. It’s anomalous, and Deep Steward doesn’t care about anomalies as much as it does similarities. It’s busy with a job, and so is the turtle, and both of their jobs are the same: examining the surrounding world, clocking what they can, plotting courses and imagining distinctions. The turtle, I have no doubt, is a turtle. But it’s also, perhaps, a Caroirendephindisplar. Who am I to tell Deep Steward, ‘no, that’s a turtle, turtle, turtle’?
The new, weird world in which we live – the uncanny landscape that surrounds us – requires a new, weird way of looking at it: landscape is a technology in and of itself, its way of seeing, and our tech needs desperately to be upgraded. We teach the machines to learn how to see, in hopes that they can begin to teach us back.
In coming months I’ll be regularly writing about works such as, and often including, Deep Steward regularly, tagged under the category of caroirendephindisplars. This new category includes uncanny landscapes, AI, video games, artworks made real, nature tourism, the post-natural, the cthulucenic and I pray many other weird taxonomies.